Biological evolution, scientific progress and the market economy run on the same engine. Something new is tried. Most attempts fail. The few that survive are kept and built upon. Genes, theories and businesses all advance this way, by trial and the elimination of error.
The obvious objection is that nature is random while science and commerce are purposive. Mutations happen by accident. Scientists and entrepreneurs act by design. Surely that changes everything?
It changes less than it seems. What is random in nature is the variation, not the selection. Mutation is blind, but the environment that judges it is anything but. It filters with relentless consistency. Selection does the real work in all three systems, and in none of them is it random.
Nor is purposive variation as sighted as it looks. Popper was clear that a scientific theory is a guess. A bold one, perhaps, an informed one, but a guess. Hayek’s entrepreneur is no different. He acts with intent but in ignorance of what customers will actually want next year. Any variation that goes beyond what we already know must be blind with respect to its outcome.
The scientist and the founder take aim, but neither can see the target. That is why most theories are refuted and most new businesses fold. Purpose does make a difference, and it makes three. It improves the odds, because educated guesses beat random ones. It allows rehearsal, because ideas can be tested in the mind before they are tested in the world. Popper said we let our hypotheses die in our stead.
A business plan can perish on a spreadsheet instead of in the bankruptcy court. Nature enjoys no such dress rehearsal. And purpose makes inheritance Lamarckian. What one generation learns, the next receives directly, without waiting for the slow lottery of the genes.
But purpose brings a danger that nature never faces. Purposive agents can tamper with the selector. A mutation cannot lobby the environment. A failing firm can lobby the government. It can seek subsidies, tariffs and licensing rules that shield it from the rivals that would otherwise replace it. An entrenched scientific school can colonize the journals and capture the grant committees. Blind variation is honest because it cannot cheat. Purposive variation can, and it tries with tireless ingenuity.
This is why a free economy and an open society need deliberate defence of competition and criticism. In biology, the selection mechanism protects itself. In markets and in science it has to be protected by rules, because the very agents it judges have every incentive to rig it.
There is a second, subtler cost. Randomness guarantees diversity. Purposive agents tend to herd. When everyone reads the same signals and chases the same fashion, variation narrows. An entire industry or discipline can march confidently in step down the same wrong road.
Now add artificial intelligence. Will it speed evolution in all three domains?
In biology it already does, but by changing the game rather than accelerating it. Directed evolution, protein design and gene editing replace blind mutation with engineered variation. Biology is being brought into the purposive camp. That is not faster natural evolution. It is the end of purely natural evolution wherever we choose to intervene.
In science, AI makes conjecture cheap and copious. It can propose hypotheses by the thousand and dispatch many of them in simulation before anyone sets foot in a laboratory. Yet the bottleneck simply moves. Reality still has the final say, and it answers at its own pace. Experiments take time. Faster guessing helps, but testing sets the speed limits.
In markets, AI slashes the cost of starting, prototyping and testing. More firms will be founded and more will fail, and they will do both faster. That is a feature, not a flaw. Rapid failure is how a market learns, and cheaper failure means more learning for the money.
The threat is the herding problem writ large. If a handful of models shape most of the hypotheses and most of the business plans, variation shrinks even as its volume soars. A million variations on a single idea amount to one idea. And the temptation to capture the selector grows with the stakes. Incumbents will be eager to help regulators decide who may use the new tools, and regulators are rarely reluctant to be helped.
So AI can turbocharge all three engines of progress. The condition is that variation stays varied and selection stays free. Nature managed both without being told. Governments, alas, will need constant reminding.
Madsen Pirie